Surprise, Surprise -- Austrian Economics, Probability and the NCAA Tournament

|Peter Boettke|

Something amazing happened yesterday -- it has never happened before in the recorded history of college basketball -- a team came back from 12 points down in only 40 seconds.  Texas A&M's basketball team was the "author" of this spectacular comeback story, and Northern Iowa went down in tragic defeat.  The particulars of this reversal of fortune -- the stakes were a trip to the Sweet Sixteen -- are best left to sports media. But what I want to highlight is that the Win Probability Calculator registered a 0 chance of winning for Texas A&M with 40 seconds left only for everyone to be taken by surprise. 

It might be objected that I am over-stressing a very basic point -- in sports, as in life, expect the unexpected.  Norte Dame won its game on a last second tip in by a freshman player whose only points in both NCAA tournament games was that tip in, making him the least likely of "Shinning Moment" heroes.  But what does all this say about the way we should think about the fluid, dynamic, and every changing reality of economic life?  Surprises do indeed happen -- the discovery of previously unrecognized opportunities are made, and those unrealized gains from trade and gains from innovation are realized only after they are recognized.

Does this mean that economists (and basketball coaches) must abandon theories of how the game will be played?  Of course not, but they should have a healthy sense of expecting the unexpected and as a result note how actors within those moments find ways to cope with such dramatic uncertainty, and how those coping strategies sometimes succeed and other times fail. 

Austrian economics is a rational actor model at its core, but rational choice as if the choosers were human.  The focus of analytical attention is on the processes of exchange and production guided by the price system, not on the equilibrium state of affairs that would result if tastes, technology and resource availability are frozen and all the gains from trade are exhausted.  That end state has a vital theoretical purpose, but it should not be the focal point of analytical attention.  Economists must avoid close-ended and single exit theories if we are to make progress in understanding the multiple margins of adjustment and adaptation to the constantly changing circumstances in a vibrant market economy.  

This is one reason I don't believe stagnation stories unless we identify and focus in that story on institutional barriers to innovation -- stunning surprises lurk right around the corner, crushing defeats are possible for those who might think of celebrating too early, and both poor execution in positions of comfort and brilliant performances amidst adversity are on display almost daily.

One of my professors -- the great Kenneth Boulding -- used to teach us that the real world was a muddle and it would a shame if our theories didn't allow us to recognize that.  In that sense we would be theoretical clear about a world that wasn't reality while claiming that our task was to get a better grasp of reality.  This is how economics could become more precise about less and less that was relevant to understanding the world.  Hayek, of course, made a similar indictment of formalistic economics as it developed in the 20th century as did James Buchanan and of course Israel Kirzner.  

Have we learned this lesson for the practice of 21st century economics?  Not yet.  Ironically, for that to happen, it may very well be the case that "What is Old Should Be New Again."

Comments

[Ryan Murphy]

The win probability would register as "100%" instead of "99.9%" I would assume the moment the model predicts a probability of one in two thousand (99.95%) that Texas A&M would win. Maybe the win probability model actually read 99.9999%, in which case it is miscalibrated. But since it may have read something like 99.95%, and given what we know about how extreme the circumstances were, I don't see any reason to believe something is amiss. "100%" is literally a rounding error.

[Peter Boettke]

Ryan,

Do you deny that there can be genuine surprises in the world?

Perhaps human's in making "rounding errors" end up being genuinely surprised and thus panic and don't even consider options on the menu that otherwise would have allowed them to weather the storm (like throwing the ball all the way downcourt as was suggested by J. Williams after the game). All Northern Iowa needed to do was to use some clock and time would have ran out for A&M, but not 4 turnovers. What is the analogy in business life? Cash flow problems with a construction company hit with weather delays that were more severe than anticipated, or poor workmanship that has a domino effect. These issues -- some of which are quantifiable risk -- are not exhausted by quantifiable risk. Similar to our ignorance in general -- we can distinguish usefully between rational ignorance (I know I don't know), ignorance (what I think I know ain't so), and utter ignorance (I don't know that I don't know) -- I think we can distinguish between risk, manageable risk, and genuine uncertainty ... and that these distinctions matter for certain explanations and not for others.

[Gene Callahan]

99.9999% is a case of completely phony "precision." These probability numbers don't mean anything more objective than what odds The person projecting the numbers thinks one ought to take in a bet. There just is no objective thing floating out there which is the "probability" of a one off event. Mises is good on this point.

[Ryan Murphy]

I believe it was very surprising. But what you seemed to be implying is that this somehow proved something.

If a model says something will happen 75% of the time, we shouldn't "want" it to always happen." The model is correct if what it claims will happen 75% of the time happen 75% of the time. Likewise, the model is correct if what it says happens 99.95% of the time happens 99.95% of the time. The 0.05% should happen 0.05% of the time.

I can take this a step further. First, theoretically I don't think sports will provide good examples of what you are talking about, because they are so artificial and countable in comparison to the real world. Second, there is a more technical reason to be skeptical of this specifically. Most would model genuine surprise as kurtosis (i.e., fat tails), but models with binary outcomes (e.g., a basketball game) are resistant to the issue. I no longer agree with Taleb on much of anything anymore, but see his technical note on the problem: http://www.fooledbyrandomness.com/binaries.pdf

[Michele Hoster]

excelent article and comments